Asset Prices in the :

Evidence from the Market.

Liam Brunt1 & Edmund Cannon2

Abstract.3

From 1770 to 1914 the British Government closely monitored the domestic grain . It collected weekly price and quantity data for all types of grain for a large number of market towns and published these so-called Corn Returns in the Gazette. We have computerised all of the published data between 1770 and 1864, totalling around 6 million data points. In this paper we describe the precise nature of data; we discuss why, when and how it was collected; we consider the accuracy and biases of the data; and we describe how we computerised the original returns. This gives a firm foundation to our own analysis of the data, and introduces the data to other potential users.

1 McIntire School of Commerce, University of Virginia. [email protected] 2 Department of Economics, University of Bristol. [email protected] 3 This project was funded by the Economic and Social Research Council under Research Grant No. R000223071. We also thank the Royal Economic Society for a Small Grant. For able research assistance we would like to thank Rob Brewer, Anna Chernova, Saranna Fordyce, Becca Fell, Ludivine Jeandupeux, Dave Lyne, Olivia Milburn, Hannah Shaw, Derick Shore, Liz Washbrook and Alun Williams. Thanks are also due to Colin Knowles for computer support and Bristol Library for exceptional service. For helpful comments and discussion we would like to thank Julia Cerutti and Lucy White. Any remaining errors are our own responsibility.

1 1. Introduction. Between 1770 and 1914 the British Government made a concerted and sustained effort to monitor the throughout England, Wales and Scotland. At its minimum, this project involved the collection and publication of weekly prices for five different for 44 counties or regions; at its maximum, the project was extended to cover weekly prices and quantities traded for seven grains for 290 towns. In total, the Government collected about 6 million data points over the period 1770 to 1914 and these data are generally referred to as the Corn Returns. We have computerised all of the data up to 1864 and are currently undertaking an extensive economic analysis to draw out the implications of this information for British economic development. The purpose of this paper is to discuss the Corn Returns as a data source. We wanted to consider all the peculiarities of the data prior to analysing them in order that our own research was well-founded. We also wanted to introduce the data set to a wider audience, in order that other researchers can confidently download the data from the ESRC data archive and undertake their own analyses. It is inevitable that such a broad and long data set varies in terms of the precise data collected and format adopted in different time periods. Hence there are a large number of issues to be discussed in order to make a comprehensive survey of the Corn Returns. However, given how frequently the basis and format of Government statistics are changed in the late twentieth century, it will come as a pleasant surprise to most economists how unchanging are the Corn Returns over a period of 135 years. In Section 2 we begin by considering the historical background to the Corn Returns – why, when, where and how they were collected and published. In Section 3 we consider the accuracy of the Corn Returns data set - both in terms of the underlying data, and our transcription of it. In Section 4 we consider potential biases in the data – arising from factors such as geography, grain quality and missing observations. In Section 5 we provide some simple data description - looking briefly at the levels, trends and cycles in prices and quantities. We include several appendices which contain more detailed information. Appendix 1 contains a full list of towns for every period.

2 2. Historical background of the Corn Returns. Between the 1690s and the 1840s the British Government attempted to regulate the domestic price of grain.4 They did this through a whole raft of legislation known collectively as the . The purpose of this regulation was two-fold. First, the Government wanted to keep up prices in order to encourage investment in and generate an increase in productive capacity.5 Second, the Government wanted to smooth domestic prices and thereby insure both farmers and consumers against excessive fluctuations. Grain was the item of the population and its demand was highly inelastic; but grain harvests fluctuated wildly from year to year, so smoothing prices was a very difficult task. When grain output and prices were about average, imports were subject to a very high (or sometimes simply banned) in order to keep up domestic prices and encourage agricultural investment. In years of plenty, the price fell very low and the Government compensated farmers by giving them an export bounty (i.e. they paid farmers to export grain in order to get it off the domestic market). In years of dearth, the price rose very high and there was no import tariff to pay (i.e. if the world supply was perfectly elastic then there was a de facto price ceiling). Regulating the domestic price was difficult in the absence of an official price series. Initially, importation and exportation were based on the honour system and hence widely abused. In order to import grain tariff-free, the captain of a vessel simply had to swear before a local worthy that the market price in the port was above the strike price; and in order to export grain and claim the bounty, the captain simply had to swear before a local worthy than the market price in the port was below the strike price. It was not unknown for a vessel loaded with grain to arrive one day (tariff-free) and depart the next day (having claimed the bounty). Eventually, the grain department at the Treasury went wildly into deficit because it had been paying out more in bounties than it had been collecting in tariffs. Parliament then acted to prevent further abuse of the system, and the solution from 1770 onwards was to compile an official grain prices series. Thus the Corn Returns were born. The precise details of how the system operated in the first few years are slightly unclear because most of the Parliamentary papers for that period have been

4 The classic account of this episode can be found in Barnes, Donald Grove, A History of the English Corn Laws, 1660-1846 (Routledge, London, 1930).

3 lost. However, we know the rough outlines for the early period and we have detailed information from 1791 onwards. Inspectors were appointed in all the major grain markets. The Inspectors collected information on the undertaken by each grain in the local market, and this information was written up into a Return which was forwarded to the Receiver of Corn Returns in London. The Receiver then calculated average prices for each county and each District (the Districts being groups of contiguous counties). The ports were then opened or closed to importation according these averge prices. The Corn Laws were always a matter of widespread concern. Farmers and landowners were interested in maintaining high prices; corn factors and industrialists were interested in allowing (free trade implied lower grain prices and hence a lower bill for industrialists). The official price series was therefore a battleground because changing the way in which the average price was calculated could affect the likelihood of the ports being opened to trade. This in turn meant that the Corn Returns were subject to widespread public scrutiny and frequent consideration was given to refining them. Even after the Government ceased regulating domestic grain prices (that is, when the Corn Laws were repealed in 1846) the Corn Returns continued to play an important economic role. Traditionally, farmers had paid a ten per cent on output (i.e. the ) to the . But both sides recognised that on marginal product discouraged farmers from raising output. Hence most were commuted by mutual consent into a “corn rent”. That is, the farmer agreed to pay to the Church each year the money equivalent of a fixed number of bushels of each type of grain. This eliminated the disincentive effects of taxation because every extra unit of output went to the farmer; but it also meant that in high price years (i.e. when the farmer had a high income) the Church received higher tax revenues.6 The rate of tax per bushel was to be set according the official price series of the Corn Returns. Recall that the Corn Returns include only domestically produced grain in their calculations. This was ideal for setting the corn rents because both farmers and the Church wanted the corn rents to reflect the price that farmers were actually receiving for their output.

5 Farmers were indeed persuaded to undertake investments which led to a permanent increase in productive capacity. For example, land was reclaimed from marshy areas by large-scale drainage projects. 6 Evans, Eric J., The Contentious Tithe (Routledge and Kegan Paul; London, 1976), 128-132.

4 In fact, much of the analysis of the Corn Returns in the late nineteenth century (such as the Parliamentary enquiries of the 1870s) was prompted by farmers and Church leaders trying to revise the way the Corn Returns were calculated, in order to improve their economic position. We can see from the foregoing discussion that the Corn Returns were the product of a peculiar set of circumstances – the desire to regulate taxation or by reference to the market price of domestically produced goods. Hence the Corn Returns have generated quite a peculiar data source. They give us the domestic prices of only those agricultural goods which were traded internationally (i.e. grain products) – but the prices are based solely on that part of the market which was not traded internationally (i.e. the grain which was produced and traded domestically). In some ways this generates a very clean data set: for example, there is no quality variation over time due to the increase in import penetration. In other ways this makes the data of limited use: for example, the Returns do not give us the average price of grain that was actually consumed, because imports are excluded. Whether or not the peculiarities of the data are a benefit or a hindrance overall, we certainly need to be aware of them when we undertake any economic analysis.

3. The accuracy of the data. We can distinguish four stages in the production of the Corn Returns data. First, the local Inspector gathered the underlying data, calculated the average for his town, and forwarded the information to the Receiver of Corn Returns. Second, the Receiver of Corn Returns inspected the data for accuracy and computed the county averages (in the early period), before sending the results to the publisher. Third, the printer set the relevant page in the London Gazette and published the newspaper for public consumption. Fourth, we transcribed data into machine- readable format and checked it. The accuracy of the final database is determined by the errors introduced at each stage of production. Let us now consider each stage in turn.

The work of the local Inspectors. The original Act of 1770 required the Justice of the Peace (JP) for each county to monitor the price of , , and beans in

5 a number of market towns.7 The precise number of towns was to be decided by the JP within certain limits. The JP was required to appoint an Inspector of Corn Returns for each town monitored; and for every return that they provided, these Inspectors were to be paid out of the County Rate.8 The JP was also to provide an official Measure of one Winchester bushel of eight gallons, in order that the returns could be made in standard units (shillings and pence per Winchester Bushel).9 These returns had to be sent to the Receiver of Corn Returns at the Treasury, who was then to publish the results in the London Gazette. By the Act of 1791 the machinery for price collection was over-hauled and made much more rigorous in several important respects.10 First, the Act listed explicitly the towns which had to be monitored in each county.11 Second, for each of those towns the local JP had to choose grain Inspectors who were “qualified persons” but were not themselves engaged in the grain trade in any way (millers, maltsters, and various other professions were explicitly excluded).12 Third, every grain factor in each of those towns was required to make a return every Monday to the local Inspector of Corn Returns, detailing his entire trade during the week and giving the total quantity traded at each price. The return had to be made under Oath and the penalty for non-compliance was £10 per week.13 Fourth, once he had received the returns from the grain factors, the Inspector forwarded this information to the Receiver of Corn Returns on Tuesay of each week. Again, the return had to made under Oath and the penalty for non-compliance was £10 per week.14 The Inspector was to be paid five shillings per return, or more if the duties of the post were particularly onerous.15 Despite the stringent requirements of the Act of 1791, there were doubts expressed about the performance of the local Inspectors – both by contemporaries and

7 10 George III, cap. 39. The precise text of the Act as it was actually passed has been lost. The closest text still extant is a copy of the Bill which came up in Parliament to be voted on. Copies of this Bill are available from the Abbot Collection. 8 The County Rate was a local tax on land which was used to finance local expenditures such as road repair. The salary cost of the Inspector of Corn Returns would have been a miniscule proportion of the County Rate. 9 As of 13th July 1827, all returns were made in shillings and pence per Imperial bushel, under 7 and 8 George IV, cap. 87. 10 31 George III, cap. 30. 11 31 George III, cap. 30, sections 45 and 61. 12 31 George III, cap. 30, section 47. 13 31 George III, cap. 30, section 50. 14 31 George III, cap. 30, section 51.

6 later researchers.16 This led Parliament to tighten up the system of local Inspectors still further in later years. For example, in 1820 JPs were given greater discretion in setting the payment for making each corn return, to ensure that they could find suitably qualified candidates; Inspectors were given the power to inspect the trade books of corn factors;17 and in 1842 the responsibility of making the returns was transferred to local excisemen who worked directly for the Inland Revenue Department of the central government.18 Even in the face of these efforts to ensure full compliance, throughout the whole period two features of the data suggest that there was substantial under-reporting of trade. First, several Inspectors made a Return of “None Sold” almost every week. This included some Inspectors in fairly large towns such as Macclesfield and Stockport, where it is hard to believe that there was genuinely never any trade in grain. In fact, this type of omission this is not really a problem: de facto, the Macclesfield and Stockport grain markets were simply not monitored and we can exclude them from the sample. This does not noticeably reduce the density or geographical distribution of the coverage. Second, some towns which furnished regular returns often seem to have had a very low level of trade. For example, the major port and market town of Bristol – with forty corn factors – often seems to have traded only around 240 bushels of wheat per week. This is an incredibly low figure, bearing in mind that wheat was the staple food grain and almost all wheat was traded off the farm and into towns. The list of monitored towns included all the major market towns of the kingdom; so overall we would expect to see almost the entire wheat crop of the kingdom passing through the monitored markets each year. In fact, we observe a volume of trade which is equal to only around 25 per cent of total domestic output. What happened to the rest, and is there a systematic difference between the grain that we observe and the rest? The most gloomy interpretation is that the grain factors or Inspectors were lackadaisical and could be bothered to return only some portion of the total trade that they undertook or observed. A more benign interpretation is that a great deal of grain

15 31 George III, cap. 30, section 71. 16 Vamplew, Wray, ‘A Grain of Truth: the Nineteenth Century Corn Averages,’ Agricultural History Review (1980), vol. 28, part 1, 1-17. 17 Evidence from Mr. William Grey Fearnside, a corn dealer in Mark Lane. British Government, ‘Select Committee on the Sale of Corn,’ BPP (1834), vol. 7, 115.

7 was traded outside the market, and hence would not appear in the returns. This could occur for perfectly innocent reasons.19 For example, we know that in the nineteenth century grain merchants sometimes bought grain “on the stalk” (i.e. before it was harvested); this transaction would not have been recorded in any of the monitored markets. Also, the market was not open every day and it is likely that corn factors were active on non-market days; probably these transactions did not form part of the return given to the Inspector. Also, if a factor was resident in one town (say ) but undertook his trade in another town (say Liverpool), then he was unlikely to be asked to make a return. From the point of view of obtaining a representative sample of prices, these practices are unproblematic. It is likely that the monitored and non-monitored grains were identical and they probably traded at the same price (it just so happened that the merchants wanted to trade on a day when the exchange was closed).20 Robert Giffen argued that an accurate estimate of the true price could be secured by sampling as little as 20 per cent of the total grain traded; the Corn Returns certainly represent more than 20 per cent of the total grain traded, so we need not worry on that account.21 One final point needs to be considered. There was some suspicion that grain factors attempted to manipulate the Returns.22 It was conceivable that grain factors could do this because some towns had very little trade, so a grain factor could substantially alter the average price by selling a large parcel of grain at an inflated price. This would obviously be of great concern to us: it would suggest that some of the prices that we observe are not true market prices, but just a fiction. Vamplew points out that under the law of 1815, if the domestic price rose above the strike price then foreign grain would be admitted free of tariff for several months – so grain factors could bring in lots of cheap foreign grain and make a large profit until the

18 Vamplew, Wray, ‘A Grain of Truth: the Nineteenth Century Corn Averages,’ Agricultural History Review (1980), vol. 28, part 1, 6. 19 For a discussion, see Fay, C. R., ‘Price Control and the Corn Averages under the Corn Laws,’ Economic Journal (1926), vol. 36, 151-2. 20 It was possible that farmers who sold grain on the the stalk received a lower price. But this would reflect both the cost of the credit being extended by grain merchant and the partial insurance which was offered (grain merchants paid per acre, with only a minor ex post adjustment in the case where the harvest was unexpectedly high or low). There is no reason to suppose that the grain itself was systematically different from the grain which was traded in the market. 21 Giffen, Robert, ‘The Gazette Average Prices of Corn,’ Journal of the Statistical Society (1879), vol. 42, 711. 22 Vamplew, Wray, ‘A Grain of Truth: the Nineteenth Century Corn Averages,’ Agricultural History Review (1980), vol. 28, part 1, 5.

8 market price was driven down to the world price. Hence there might be both a means and an incentive to inflate the average price. There are three major objections to this argument. First, there is no evidence that this ever actually occurred – it was just rumoured that it might do so. Second, it is true that some factors may have wanted to import cheap foreign grain. But the factors who were holding British grain most definitely did not want to allow any imports, because it would drive down the price of the assets which they were holding. The average level of imports over the period was around 10 per cent of domestic output - so the vast majority of grain factors had their money invested in domestic grain, rather than foreign grain. So if there was any manipulation then it seems more likely that the holders of British grain would be working to exclude imports by artificially reducing the averages (i.e. by selling under-priced parcels of grain in small markets). No one has ever suggested that this occurred. Third, manipulating the market price was not as easy as one might think. The averages include only those transactions in which grain physically changes hands. The factor was allowed to include a transaction in his return to the Inspector only once delivery of the grain had been made. In that sense all the transactions that we observe are bona fide, and not just paper transactions.23

The work of the Receiver of Corn Returns. In the period up to 1820 the Returns reported an average price for each county. The Receiver was allowed to publish an average price for a particular county for a particular week if and only if he received price data from at least two thirds of the towns monitored in that county.24 All prices were to be calculated as weighted averages - whether they were prices for towns, counties or districts.25 Through most of 1796 the newspaper Bell’s Weekly Messenger reported the official average wheat price for the Mark Lane grain market in London, as well as a breakdown of the quantities traded at each price. Hence we were able to verify that weighted averages had indeed been used in calculating the official average.

23 Evidence of Mr. William Gray Fearnside: British Government, ‘Select Committee on the Sale of Corn,’ BPP (1834), vol. 7, 114. 24 31 George III, cap. 30, section 67. 25 31 George III, cap. 30, section 52. Although weighted average prices were published for each county, importation in this period was actually determined by the weighted average prices of the 12 Districts. 31 George III, cap. 30, section 54.

9 Increased professionalisation at the local level was matched at the upper levels of the organisation. From 1820 the returns were published by town, rather than appearing as county averages. We then get direct evidence that all the returns were checked before publication, since some of the town returns are witheld as being an “Incorrect Return”. From 1822 to 1842 William Jacob was Receiver of Corn Returns.26 Jacob was an absolute expert on the grain trade. He started out as a London merchant before becoming a Member of Parliament; he had travelled widely in Europe and collected a huge amount of evidence on the international grain trade, about which he wrote prolifically.27 He provided much of the evidence considered by Parliamentary Committees when the Corn Laws were reformed. The machinery could not have been run by a more suitable candidate. Jacob’s successors were equally competent. One who stands out is Robert Giffen, who became Comptroller of Corn Returns in 1876. The accuracy of the Returns was questioned in the late 1870s because they seemed to indicate a decline in domestic production. Numerous potential weaknesses of the Returns were highlighted, particularly the fact that various provincial markets persisted in using local measures for trade (which then had to be converted into Imperial units). This critique undermined confidence in the price data as well as the quantity data. Giffen made an exhaustive investigation of the measurement issue and concluded that in the worst-case scenario (i.e. where the proper conversions had not been made) the Returns would over-estimate the true average price of wheat by 1.25 per cent.28

The work of the publisher. The final link in the information chain comes with the publisher. No matter how accurately the returns were calculated by the Reciever of Corn Returns and his staff, everything comes to nought if the publisher failed to

26 Goodwin, Gordon, ‘Jacob, William,’ in The Dictionary of National Biography, ed. Sidney Lee (Smith, Elder and Company; London, 1896), vol. 26, 122-3. 27 For example, see William Jacob, Considerations on the Protection Required by British Agriculture and on the Influence on the Price of Corn on Exportable Productions (8 vols., London, 1814); and his Report on the Trade in Foreign Corn, and on the Agriculture of the North of Europe (2nd ed., 8 vols., London, 1826); and his most famous work Tracts Relating to the Corn Trade and Corn Laws (8 vols., London, 1828). 28 Giffen, Robert, ‘The Gazette Average Prices of Corn,’ Journal of the Statistical Society (1879), vol. 42, 717. This issue had already been addressed in several previous Parliamentary enquiries, with the same result. For example, comparisons of regional (‘customary’) measures, and their relationship to

10 render the figures accurately. In this we are fortunate because the Corn Returns were not at the mercy of a commercial newspaper publisher. The Corn Returns appeared in the official Government newspaper called the London Gazette. The London Gazette was started in the seventeenth century and remains in print to this day. Its main purpose was to convey official information accurately: it did not have to meet commercial profit targets, and it did not hurry into print with up-to-the-minute news. The London Gazette carried official reports on military engagements (of which there were many up to 1815); and official appointments (civil and military). It also carried a lot of commercial and numerical information. There was information about banks; bankruptcies; government procurement contracts; auctions of newly-arrived exotic items (tea and such like); and many other items. Initially the Corn Returns were published with a delay of two weeks (i.e. the returns for the week ending Saturday 5th January 1771 appeared in the London Gazette of Saturday 19th January 1771). The publication process was accelerated during May 1772. The publication delay was reduced to one week for the English returns (i.e. the returns for the week ending Saturday 23rd May 1772 appeared in the London Gazette of Saturday 30th May 1772).29 The returns for Wales and Scotland continued to appear with a delay of two weeks until 6th June 1789, when they were synchronised once again with the English Returns. On one occasion an Errata was published in the London Gazette. It noted that the figures for three towns had been rendered incorrectly in an earlier week, and the correct figures were given.30 This strongly suggests that the published figures were checked for accuracy and corrected where necessary. All the available evidence suggests that the London Gazette provided a faithful reproduction of the figures provided by the Receiver of Corn Returns.

The work of the researcher. The most difficult data to transcribe was that from 10th November 1770 (the beginning) to 8th March 1823. The print quality of the London

standard measures can be found in British Government, ‘Select Committee on the Sale of Corn,’ BPP (1833), vol. 7, 54-6. 29 This acceleration occurred in two stages. The delay for English returns was initially cut to 10 days by bringing forward publication of the Returns to the Tuesday edition of the Gazette. The delay was then cut by another three days by bringing forward the publication of the Returns to the Saturday edition once again. 30 See the London Gazette for 20th November 1824.

11 Gazette was mediocre and has suffered significantly with the passage of time. First, the publisher used a very small type which is tiring to read for prolonged periods (something like a Times New Roman font in 7 point). Second, the quality of the print was not good (often rather blotchy). Third, the quality of the paper was poor and much degraded with time (there was a lot of bleed-through from the next page which sometimes made the numbers difficult to read). All of this data had to be transcribed manually by a typist. This process was rather slow (averaging around 2000 digits entered per hour, including associated data management tasks) but extremely accurate. We randomly double-checked 2 weeks (after the checking process described below) and found the data entry to be 100 per cent accurate. From 15th March 1823 onwards the publication quality was very good. Each page became less cluttered because the Receiver of Corn Returns stopped reporting regional sub-totals, and this allowed the publisher to use a much larger font. The returns for 150 towns were spread over four whole pages; and after 29th April 1842 the returns for 290 towns were spread over eight whole pages. The quality of printing and paper were both much higher and there were no problems of legibility. This enabled us to enter the data using a scanner and Optical Character Recognition (OCR) software. Perfecting the set-up of the OCR software was quite an art, and we have described this process in detail elsewhere.31 Overall, scanning was much faster than typing (averaging around 5000 digits entered per hour, including data management tasks) but rather less accurate. We randomly double-checked 2 weeks (after the checking process described below) and found the data entry to be 97 per cent accurate. The inaccuracies arise because the OCR software sometimes mis-recognises digits, usually if the printing is blotchy or the paper is flecked. Typical errors are the numerals 3 or 6 being mistaken for the numeral 8. Once the raw data had been entered we undertook a number of checks to eliminate any transcription errors. First, up to 1820 the published data is simply a price per bushel, measured in shillings and pence. Graphing the data for each year for each county allowed us to pinpoint any suspicious jumps in the price. Notably, an error in entering the value of shillings typically leads to a jump in price of 12, 24 or 36 pence (since there are 12

12 pence to a shilling). Since the average price of wheat up to 1820 was only around 100 pence per bushel, a jump of 12 or 24 pence in one week was rather noticeable. Hence any transcription errors which we have failed to notice are likely to come from the pence column of the original data sheets – and therefore they are likely to be small. Second, after 1820 the published data records the total value of each grain traded in each town, and the total quantity traded in each town. Most of the data was scanned from photocopies and many of the photocopies were very dark (sometimes illegible) along the edge which had been bound into the spine of the volumes. Hence for every page for every week we checked the rows closest to the spine to ensure that they had been scanned correctly. We also checked that all the pages were in the correct order (sometimes the pages had been bound together in the wrong order and this was not always apparent at first glance). We then calculated the average price per unit by dividing the total value traded by the total quantity traded. We then wrote a programme which highlighted the highest and lowest prices recorded each week for each type of grain: these were typically erroneous and had to be corrected. It would obviously be desireable to have 100 per cent accuracy in data transcription. We could achieve this (or come very close to it) if all of the data were entered manually and then checked by graphing. However, this would be extremely time-consuming and it is doubtful if the increase in accuracy would justify the additional cost. An alternative way of thinking about the problem is that we must choose between entering 100 per cent of the data with 97 per cent accuracy and entering 40 per cent of the data with 100 per cent accuracy. It seems clear that the precision of our estimates (for example, using regression techniques) will be higher if we collect 100 per cent of the data with 97 per cent accuracy. Hence we believe that we have the optimised the data collection process and chosen the best course to steer between two pitfalls.

4. Biases in the data. There are potentially three sources of bias in the Corn Returns data. First, there could be bias in the geographical coverage (certain regions or types of regions may be over- or under-represented). Second, there could be bias in the temporal coverage (certain years, or periods within each year, could be over- or

31 Jeandupeax, Ludivine, ‘OCR Packages,’ mimeo., University Of Bristol. After extensive testing of several packages we found that Abbyy Finereader (correct spelling!) was by far the most accurate and

13 under-represented). Third, there could be bias in the quality of the grains traded (high or low quality grains may more likely to be traded). We should also note that these biases need not be inherent in the system – they might arise from the way that the system operated. In particular, there are quite a few missing price observations which are not randomly distributed, and we need to consider their impact on the representativeness of the data. We now consider each of these issues in turn.

Geographical coverage. We would like to have a fairly even geographical coverage, for two reasons. First, we would like to know what was happening to market prices in every region in order to link that information to other economic data (so that we can gauge the effect of local transport and financial structures on prices and so on). Second, grain production was spread quite evenly across the country, especially before 1840, so we need to monitor every region in order to know what was happening to grain production in aggregate. This uniformity was driven partly by high transport costs, which made it efficient to produce bulky grain products locally even if the local climate was poorly suited; and it was driven partly by the organic nature of farming, which required a mixture of crop and animal production. The geographical coverage of the towns was very good between 1791 and 1820. A large number of towns were monitored (213) and they were scattered fairly evenly across the country, so that at least one town was monitored in every English and Welsh county. This can be seen in Table 1 below, where we have calculated the number of monitored towns per 1000 square miles. At first glance, many of the numbers seem to be rather small – two towns per 1000 square miles does not sound like a very dense coverage. However, if you took a town and drew around it a circle with a radius of 12.6 miles, then the catchment area within the circle would be 500 square miles. So a density of two towns per 1000 square miles implies that (very roughly speaking) everyone lived within 12.6 miles of a monitored town. This is approximately the distance that farmers took their grain to market in 1770.32 However, Parliament decided in 1804 that the importation of corn should be regulated by the price of domestic corn in maritime districts only. Public interest in the inland counties therefore waned and when the list of towns was revised in 1821

efficient package. 32 Young, Arthur, The Farmer’s Tour through the Eastern Counties of England (London, 1770).

14 (down to 139) all of the towns in inland counties were removed. The coverage of the Corn Returns was then highly skewed and no data was collected for the 20 inland counties (comprising 28.8 per cent of the total land area of England and Wales). Again, this is very clear in Table 1.

Table 1. Monitored Towns per 1000m2. County 1791- 1821- 1828- 1842- 1864- 1883- 1890- 1900- 1820 1828 1842 1864 1883 1890 1900 1912 Beds 6.5 0.0 2.2 6.5 2.2 4.3 4.3 4.3 Berks 4.3 0.0 2.8 8.5 2.8 8.5 8.5 7.1 Bucks 4.1 0.0 1.4 5.5 1.4 2.7 2.7 2.7 Cambs 3.7 3.7 3.7 3.7 3.7 3.7 3.7 3.7 Chester 3.6 3.6 3.6 6.3 3.6 0.9 0.9 0.9 Cornwall 4.4 4.4 4.4 7.3 5.1 2.2 2.2 2.2 Cumberland 2.6 2.6 3.2 5.1 3.2 1.9 1.9 1.9 Derbs 2.9 0.0 1.0 1.9 1.0 1.9 1.0 1.0 Devon 2.3 2.3 2.3 3.5 2.3 2.3 3.5 3.1 Dorset 6.1 6.1 6.1 7.1 6.1 5.1 5.1 5.1 Durham 6.2 6.2 6.2 6.2 6.2 4.1 4.1 4.1 Essex 1.8 1.8 1.8 3.6 0.6 3.6 3.0 3.0 Gloucs 4.0 4.0 4.0 7.2 4.0 2.4 3.2 3.2 Hants 5.4 5.4 5.4 6.0 5.4 4.2 4.2 4.2 Hereford 3.6 0.0 0.0 3.6 0.0 1.2 2.4 2.4 Herts 4.9 0.0 3.3 9.8 3.3 6.5 6.5 6.5 Hunts 8.4 0.0 2.8 5.6 2.8 8.4 8.4 5.6 Kent 1.8 1.8 1.8 4.3 2.5 4.3 4.3 3.7 Lancs 4.2 4.2 4.2 5.8 4.2 2.6 2.1 2.1 Leics 3.7 0.0 1.2 5.0 1.2 3.7 3.7 3.7 Lincs 2.9 2.9 2.9 6.5 2.9 3.2 3.2 3.2 Mids 10.7 0.0 3.6 3.6 3.6 3.6 3.6 3.6 Mons 6.9 6.9 6.9 8.7 6.9 5.2 5.2 5.2 Norfolk 5.7 5.7 5.7 6.1 5.7 4.7 4.7 4.7 Northants 3.0 0.0 1.0 5.1 1.0 3.0 4.1 3.0 Northumberland 3.1 3.1 3.1 3.1 3.1 1.5 1.5 1.5 Notts 3.6 0.0 2.4 4.9 2.4 4.9 6.1 6.1 Oxon 4.1 0.0 1.4 6.8 1.4 4.1 4.1 4.1 Ruts 13.4 0.0 0.0 6.7 0.0 6.7 6.7 6.7 Salop 2.3 0.0 0.0 5.4 0.0 2.3 3.9 3.9 Somerset 3.7 3.7 3.7 6.1 3.7 3.7 3.7 3.7 Staffs 2.6 0.0 0.0 7.9 0.0 1.8 2.6 2.6 Suffolk 6.1 6.1 6.1 6.8 6.1 7.4 8.8 8.8 Surrey 4.0 0.0 1.3 5.3 1.3 6.7 6.7 6.7 Sussex 2.0 2.0 2.0 6.8 2.0 3.4 4.1 4.1 Warks 3.4 0.0 2.3 4.5 2.3 4.5 4.5 4.5 Westmoreland 2.6 2.6 2.6 2.6 2.6 1.3 1.3 1.3 Wilts 2.2 0.0 0.7 4.4 0.7 3.7 3.0 3.0 Worcs 4.1 0.0 1.4 6.8 1.4 1.4 2.7 2.7 Yorks 1.7 1.2 1.7 4.0 1.7 3.5 3.2 3.2 London 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 N Wales 5.5 4.8 1.0 2.9 1.0 1.3 1.6 1.6 S Wales 3.6 2.7 0.7 1.8 0.7 1.1 1.1 0.9

15

The situation improved somewhat between 1828 and 1842, when the number of Welsh towns was slashed and a number of towns in inland counties were added instead. Only four counties were entirely unreprepresented in this period (Hereford, Shropshire, Staffordshire and Rutland) which together comprised only 6.7 per cent of the land area of England and Wales. The returns were at their most comprehensive between 1842 and 1864 when the number of towns rose to 290 and all counties were again represented. Between 1864 and 1883 the number of towns was once again reduced to 150 and the geographical coverage was almost exactly the same as the period 1828 to 1842 (only one or two towns differed). Then the situation improved once again from 1883 to 1914 – the number of towns was increased slightly and the geographical distribution made more even – so that coverage was similar to the period 1771 to 1820. Our survey shows that the geographical coverage of the Corn Returns was generally good between 1770 and 1914. Over that period of 135 years, returns were made for all counties in 94 years; and returns were made for all but four counties in 34 years. The only really deficient period occured in the seven years from 1821 to 1828, when there was no coverage of the inland counties.

Temporal coverage. The temporal coverage of the Corns Returns is both very extensive and very consistent. We have seen that the Returns run from 1770 to 1914; and the breadth of coverage through the period is variable but almost always good. The Returns run weekly throughout every year. The volume of trade for most crops is fairly constant through the year, so there is no a priori reason to treat particular phases or sub-periods with undue suspicion. The volume of trade in barley tends to be somewhat concentrated in the autumn and early winter, when maltsters were trying to secure the best quality barley for beer production. But there is still a fair quantity of trade through the rest of the year.

Grain quality. Grain varies greatly in quality. When a parcel of grain is milled it produces several different types of , which are ranked according to their fineness. Suppose that we take a bushel each of high and low quality grain and mill them into flour. Then the low quality grain will produce fewer ounces of flour in total, and a

16 higher proportion of that flour will be coarse. Each ounce of coarse flour will produce fewer ounces of (as compared to the fine flour); and each ounce of that bread will have a lower value (as compared to the bread made of fine flour). That is because - until recently - consumers required a price discount in order to persuade them to eat brown bread. The value of a bushel of grain is determined by the value of the bread which is produces; so a lower quality grain (which produces a lower value of bread) will sell for a lower price per bushel.33 For each type of grain there was a quality differential between British and foreign produce; and there considerable quality variation within the domestic product. This is likely to affect the grain price data in four important ways. First, there will be a differential between the levels of British and foreign grain prices. We noted above that the Corn Returns were designed to monitor the domestic price of domestically produced grain only: all foreign grain was specifically excluded. Imported grain was typically of lower quality and traded at a lower price than domestically produced grain.34 For this reason, the Corn Returns over-estimate the average price of grain which was actually consumed in England, Wales and Scotland. Moreover, since imports were rising as a proportion of consumption the strength of this effect was rising over time (i.e. there was an increasing divergence between the average price of grain in the Corn Returns and the average price of grain that was actually consumed). This caveat could prove to be important, for example, if grain prices were used to calculate long run changes in the cost of living. Since we have good data on the price of imported grain, it would actually be straightforward to control for this effect. Second, there was regional variation in grain quality which would affect the levels of local grain prices. For example, in Lincolnshire the farmers grew a variety of wheat called Rivetts; this was suited to the Lincolnshire climate and had a high yield per acre, but it sold for a lower price per bushel because it was not good for bread- making.35 This type of variation would generate differences in grain price levels across regions even in the face of perfectly efficient markets with zero transport costs.

33 British Government, ‘Select Committee on the Sale of Corn,’ BPP (1833), vol. 7, 251. 34 Prices of imported grains can be found in various contemporary sources, such as Bell’s Weekly Messenger. These sources almost invariably show that imported grain traded at a lower price per bushel. 35 Percival, John, Wheat in Great Britain (Duckworth, London, 1934), 69-73.

17 Currently, there is no systematic data available on regional wheat varieties before 1914 so it would be difficult to control for this effect explicitly. Third, the average quality of grain varied markedly from year to year, and we would expect this to significantly affect the average level of prices from year to year. However, this will not be apparent from looking at the raw data because there is a confounding effect. Roughly speaking, bad weather generates a harvest in which the grain is low in quality and small in quantity. The low quality of the grain would put downward pressure on prices per bushel; but the small quantity of the grain would put upward pressure on prices per bushel (it is an adverse supply shock). Since the demand for grain was very inelastic (it was the staple food product) the quantity effect on price per bushel will outweigh the quality effect. But an important implication is that if we use time-series data to estimate the elasticity of demand for grain then we will over-estimate the elasticity (because we would under-estimate the price increase in years of dearth). A second point is that if we calculate the output of bread based on the output of grain then we need to control for year-on-year variations in the quality of flour. There is good data available concerning the effect of grain quality on bread- making, so it would be straghtforward to control for these effects. Fourth, it might be the case the cycle of grain prices within the year was be affected by grain quality. There is always some low quality grain produced in every year and in every locality. Every field of grain produces some low quality output called “tail corn”, which constitutes around ten per cent of total grain output. Low quality grain has the much same weight and volume as high quality grain (in the early 1800s, low/high quality grain weighed around 58/62 pounds per bushel). It therefore cost about the same to send low or high quality grain to market – but the low quality grain had a lower value per bushel. Therefore it was less profitable to send the low quality grain to market and we systematically observe trade in higher quality grains. Low quality grains were kept on the farm for feeding farm servants or fattening animals for market. However, suppose that prices were unexpectedly high later in the year. Then it would be profitable to send to market the low quality grain which was expected to be consumed on the farm. Then low quality grains would systematically come onto the market later in the year, which in turn would put downward pressure on the upswing in prices. We would then under-estimate the upswing in prices.

18 One further point is worth noting. We know that it was often not worth sending the lowest quality grain to market; this put upward pressure on average prices. But we also know that the best quality grain did not go to market either. The best grain was used as corn and was traded directly between farmers. It would never go via a grain factor and enter into the Corn Returns. This put downward pressure on average grain prices. Overall, we can say that the variation in grain quality was substantial and fluid over time. It is likely to be an issue which needs to be considered carefully in any use of the Corn Returns or other grain prices. Otherwise we might well draw erroneous or inaccurate conclusions.

Missing observations. Let us first consider the period after 7th October 1820, when the Returns for each individual towns are published (rather than just county averages). Price observations can be missing from the Corn Returns for four reasons, and the cause of each missing price observation is described according to the following system. First, it might be the case that no Inspector of Corn Returns was appointed for a particular market. This is recorded in the London Gazette as ‘No Inspector’ and occurred very rarely. In the period 1820 to 1825, the only towns to be found without an Inspector were Swansea (20th and 27th April 1822) and Kidwelly (17th November 1821 to 30th March 1822). Second, the local Inspector of Corn Returns might fail to send in a return. This is recorded as ‘No Return’. Third, the Inspector might send in a Return which was incorrect. If the whole Return was incorrect then this is recorded as ‘Incorrect Return’; if it was just one or two crops then ‘Incorrect’ is entered for that particular column or columns.36 It is not clear how the office of the Comptroller of Corn Returns knew that the return was incorrect. Probably the return was internally inconsistent (i.e. the volume traded multiplied by the price per unit did not equal the value traded). For one week only, the

36 In fact, the use of ‘Incorrect’ for individual crops begins on 8th September 1821. Prior to that date we only ever observe ‘Incorrect Return’. It seems likely that prior to 8th September 1821 the Receiver discarded the whole Return if even one of the crops was incorrect; this was perhaps unnecessarily harsh, and hence the policy was changed. The effect of this change in policy is clear from Table 2 - there is a discrete drop of around 80 per cent between 1821 and 1822 in the number of missing observations due to ‘Incorrect Return’.

19 legend ‘Irregular Return’ is employed (on 23rd December 1820). It is not clear how this is different to an ‘Incorrect Return’, and we have treated them in exactly the same way. Fourth, it could be the case that there was no trade. If there was no trade in any crop then this is recorded as ‘None Sold’; if it was just one or two crops then a dash is entered for that particular column or columns ‘ – ’. If there was no trade then we can enter a zero for the quantity. But in the absence of any trade, we do not observe a price and hence we have a missing price observation. In the early months of 1821 the legend ‘None Brought for Sale’ was employed (see, for example, 31st March 1821). Presumably, the use of ‘None Sold’ then indicated a situation where grain was brought for sale but there were no buyers. However, this distinction was soon dropped and only the legend ‘None Sold’ appeared thereafter. We have treated the two cases in exactly the same way. In our spreadsheets of grain prices we do not describe why the price observation is missing. However, we may wish to know why the price data is missing - whether it was due to an absence of trade or poor reporting - and we have recorded this fact in an additional spreadsheet. For the period 1820 to 1825, Table 2 below shows the number of prices which are missing each year owing to there being: no trade; having something ‘Incorrect’; or there being ‘No Return’.37 Table 3 below shows the distribution of missing observations across crops.

Table 2. Breakdown of Missing Price Observations by Cause of Omission, 1820-25. Year ‘No Return’ ‘Incorrect’ or ‘None Sold’ or ‘Incorrect Return’ ‘ – ’ 1820 445 (5.4) 850 (10.3) 6953 (84.3) 1821 1220 (5.0) 1969 (8.2) 20970 (86.8) 1822 1245 (6.1) 384 (1.9) 18942 (92.0) 1823 1825 (7.8) 259 (1.1) 21265 (91.1) 1824 315 (1.6) 166 (0.8) 19562 (97.6) 1825 280 (1.5) 107 (0.6) 18255 (97.9) Note. (a) Figures in parentheses give each type of omission as a percentage of the annual total. (b) The data for 1820 run from 7th October only.

37 Note that a ‘No Return’ or an ‘Incorrect Return’ generates five missing price observations (wheat, barley, oats, beans and peas), whereas each ‘Incorrect’ generates only one missing price observation.

20 Table 3. Breakdown of Missing Price Observations by Crop, 1820-25. Year Wheat Barley Oats Beans Peas 1820 371 (4.5) 990 (12.0) 1427 (17.3) 2598 (31.5) 2862 (34.7) 1821 2102 (8.7) 3672 (15.2) 4687 (19.4) 6909 (28.6) 6789 (28.1) 1822 1296 (6.3) 2859 (13.9) 4361 (21.2) 5534 (26.9) 6521 (31.7) 1823 1144 (4.9) 2685 (11.5) 4833 (20.7) 7985 (34.2) 6702 (28.7) 1824 661 (3.3) 2265 (11.3) 3688 (18.4) 5572 (27.8) 7857 (39.0) 1825 727 (3.9) 1566 (8.4) 3150 (16.9) 5443 (29.2) 7756 (41.6) Note. (a) Figures in parentheses give each type of omission as a percentage of the annual total. (b) The data for 1820 run from 7th October only.

A detailed look at the data suggests that the pattern of missing observations is much as we would expect. Most of the missing price observations (90 per cent) arise from an absence of trade. It is also more common for the prices of the minor crops to be missing, because it was more likely that there would be zero trade for a minor crop in any particular week. Overall, there are very few missing observations for the major crops and the major towns. By contrast, the proportion of missing observations is very high for some towns, and it is doubtful whether it is worthwhile to attempt a systematic analysis of what remains for those towns. In the period 1770 to 30th September 1820, the grain prices are published as averages for each county. In this period the cause of a price observation being missing is not revealed. We noted above that the Receiver would not publish a county average if fewer than two thirds of the Inspectors made a return. Presumably the Returns from various Inspectors within a county could be missing for a variety of reaons, so it would probably not be straightforward to reveal why a price observation was missing. Given the probability of each type of missing observation after 1820, the most likely explanation for a missing observation in the period before 1820 is simply ‘None Sold’. In Table 4 below we show how much data we have for each crop for English and Welsh counties (expressed as a percentage of the total possible amount of data, given the number of counties and the number of weeks between 1770 and 1820). Notice that the figure for beans noticeably lower than the rest. This is because bean prices are basically never reported in Cornwall, Devon, Derbyshire and Cheshire; this is probably because very few beans were produced in those counties and hence there was virtually no trade. By contrast, bean prices are reported almost every week in the other counties.

21 Table 4. Reporting Rates for Grain Price Data, 1770-1820. 40 English Counties 2 Welsh Regions Wheat 99.7 93.0 Barle 93.1 92.6 y Oats 97.2 88.0 Beans 79.0 15.7

One further point is worth noting with regard to the period up to 1820. The practice of publishing a county average even if one third of towns did not make a Return could generate measurement error. This is because the county average might not be based on a constant sample of towns. There is no way that we can further illuminate this issue. In general, the proportion of price data which is missing from the Corn Returns is very low. Moreover, it is heavily concentrated in a few remote towns, such as Lampeter. We can afford to exclude those towns from our analysis (the region as a whole is adequately covered by a good set of Returns for other towns). Hence there is no problem at all with missing observations.

5. Description of the data. There are four main aspects of the data that we might like to consider. First, we can look at the long run movements in the price of each grain. Second, we can look at changes in relative prices between grains. Third, we can look at the short period pattern of prices. Fourth, we can consider the regional variation in price levels. In Figure 1 below we graph the long run price movements for wheat. The grey line is the weekly national average price (a simple mean of all the counties); the red lines show the highest and lowest county prices each week to demonstrate the price range. It is clear that all of the county prices bascially move together over time. This is not surprising, since they are all affected by common shocks (be they supply shocks from the weather, or demand shocks from wars and such like). These long run price movements are already well-known - the annual averages of the grain prices published in the London Gazette are reproduced in the Parliamentary Papers and have been widely used by researchers.

22 Figure 1. The Long Run Price of Wheat, 1770-1820 (d/quarter).

350

300

250

200

150

100

50

0 1770 1773 1775 1778 1780 1783 1785 1788 1790 1793 1795 1798 1800 1803 1806 1808 1811 1813 1816 1818

In Figure 2 below we graph the price of each grain relative to the price of wheat. It is apparent that there is no long run trend in relative prices over the period 1770 to 1820, although there are short term fluctuations. On implication of this consistency in price relatives is that the long run movement of the minor grains tracks the price of wheat (and hence the long run graphs for barley, etc. look basically the same as Figure 5 above).

23 Figure 2. Relative Prices of Wheat, Barley, Oats and Beans, 1770-1820.

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 Price Barley / Price Wheat Price Oats / Price Wheat 0.1 Price Beans / Price Wheat 0 1770 1773 1776 1778 1781 1783 1786 1789 1791 1794 1797 1799 1802 1804 1807 1810 1812 1815 1818 1820

In Figure 3 below we present three sample wheat price series drawn from different English counties. The vertical lines on the graph divide the data approximately into harvest years. The cycle of prices within each year is evident. The wheat harvest in most English counties occurs around the last week of August (with a standard deviation of two weeks).38 Prices then typically reach a low point around the end of October (i.e. when the new harvest starts coming onto the market in bulk). Thereafter, the wheat price rises until sometime in August and the arrival of the next harvest. In Figure 4 below we present analagous data for prices; the price cycle is again apparent.

38 Lawes, J. Bennet, and J. H. Gilbert, ‘Report of Experiments on the Growth of Wheat for 20 Years in Succession on the same Land,’ Journal of the Royal Agricultural Society of England (1864), vol. 25, 93-185.

24 Figure 3. County Wheat Prices, 1779-84 (d/quarter).

100

Durham Berks Price Lancs

40 06/11/1779 05/11/1780 04/11/1781 02/11/1782 01/11/1783 31/10/1784 Year

Figure 4. County Oat Prices, 1781-89 (d/quarter).

45

Cornwall Cumberland Price Durham

15 02/11/1782 01/11/1783 30/10/1784 29/10/1785 28/10/1786 Date

Notice that the price cycle is not always completely well-behaved. For example, if the harvest is poor then there may be little or no fall in prices between one year and the next; this behaviour can be seen in the wheat price data in 1781/2, and in the oat price data in 1784/5. Alternatively, there is sometimes little or no appreciation in prices through the year (or even a persistent price decline); this can be seen in the Berkshire wheat series in 1784.

25 6. Conclusion. The Corn Returns are the most important data source on British grain markets in the period 1770 to 1914. In fact, they are probably the largest single body of data on British agriculture before the twentieth century. This paper summarises our efforts in transcribing that data into a machine-readable format. It also considers the weaknesses and potential biases of the Corn Returns and evaluates them as a source of economic information. Overall, the Corn Returns furnish us with very good data. They are of high quality - being based on extensive underlying grain returns, and processed by a competent and thorough administration. They have a broad and fairly even geographical coverage of England and Wales (rather less for Scotland). They have an excellent temporal coverage, the data being both long run and frequent (i.e. weekly observations). There are some issues which need to be borne in mind when drawing inferences from the price data, such as the variation of grain quality across space and time. But we generally have enough information available to control for these biases. Hence the Corn Returns should prove to be a very firm basis for economic analysis and help us to gain a better understanding of British economic development during the Industrial Revolution.

26 Appendix 1. Market Towns Reported in the London Gazette. The number of towns monitored for the Corn Returns varied over time, as detailed in Table A1 below. The full list of towns for each period is reported in Table A2 overleaf.

Table A1. The Number of Towns Monitored for the Corn Returns, 1770-1914. From To No. of Parliamentary Act (Order of Council) Towns 1770 5/11/1791 10 George III, cap. 39 12/11/1791 30/9/1820 213 31 George III, cap. 30 7/10/1820 17/2/1821 210 31 George III, cap. 30 24/2/1821 31/3/1821 210 31 George III, cap. 30 7/4/1821 23/6/1821 210 31 George III, cap. 30 30/6/1821 7/7/1821 210 31 George III, cap. 30 14/7/1821 11/8/1821 139 1 and 2 George IV, cap. 87 18/8/1821 16/2/1822 148 1 and 2 George IV, cap. 87 23/2/1822 2/3/1822 149 1 and 2 George IV, cap. 87 9/3/1822 7/6/1823 149 1 and 2 George IV, cap. 87 14/6/1823 6/7/1827 150 1 and 2 George IV, cap. 87 13/7/1827 4/7/1828 150 7 and 8 George IV, cap. 87 11/7/1828 22/4/1842 150 9 George IV, cap. 60 29/4/1842 1864? 290 5 Victoria cap. 14 1864 7/4/1883 150 27 and 28 Victoria, cap. 87 14/4/1883 5/4/1890 187 45 and 46 Victoria, cap. 37 (OC 14/2/1883) 12/4/1890 31/12/1900 196 45 and 46 Victoria, cap. 37 (OC 12/3/1890) 1/1/1901 31/12/1912 190 45 and 46 Victoria, cap. 37 (OC 26/11/1900) 1/1/1913 31/12/1914 173 45 and 46 Victoria, cap. 37 (OC 11/10/1912) Notes. (a) It is commonly suggested that the returns for the period 1791 to 1820 are based on 212 towns in England and Wales (for example, BPP 1890-91, vol. 65, 165). However, the list of 212 towns in the Act makes no mention of London – whereas data was definitely collected for London throughout the period. London prices are reported in the London Gazette from 1770 to 1795; and thereafter, London prices and quantities traded are reported in other publications, such as Bell’s Weekly Messenger. So the total number of towns was actually 213. (b) The Corn Returns Act of 45 and 46 Victoria, cap. 37 enacted that: “Weekly returns of the purchases of British corn should be made under the direction of the Board of Trade in manner provided by this Act from such towns, not less than 150 and not more than 200 in number, as may from time to time be fixed by Her Majesty in Council, and the average price of British corn shall be from time to time ascertained from those returns, and published by the Board of Trade in manner provided by this Act”. See BPP 1888, vol. 10, 140. OC=, in the above table.

Nomenclature. The nomenclature of towns is somewhat wayward in the returns published in the London Gazette. Two particular issues arise. First, the spelling of town names is sometimes incorrect and sometimes inconsistent. For example, initially returns are reported for the town of Tunbridge in Kent; latterly this changes to Tonbridge. In fact, there is no town of Tunbridge in Kent – there is a town called Tunbridge Wells and another called Tonbridge. Given

27 the change in spelling to Tonbridge, it seems likely that the returns which were being reported in the early years were actually gathered from Tonbridge. We have assumed this to be the case and we use the name Tonbridge throughout. Second, the names of some towns changed over the period. For example, initially returns are reported for the town of Glanford Brigg in Lincolnshire (also Glandford Briggs and Glandford Bridge!); latterly this changes to Brigg. These names definitely refer to the same town throughout because we can identify the location on maps from the eighteenth and nineteenth centuries. A more radical change occurred for Shaston in Dorset, which is now known as Cann. In order to prevent any ambiguities we have adopted modern names and spellings for all the towns throughout the period 1770 to 1914.39 With the exception of Brigg and Cann, the modern spellings are reasonably close to those used in the original returns. Hence anyone using our data set will find it straightforward to match up the lists of towns which we report and the lists reported in the London Gazette (there is certainly no need to reproduce a key in this paper).

Bristol. In the eighteenth and early nineteenth centuries Bristol was an important commercial city with a thriving grain market, lying exactly on the borders of Somerset and Gloucester. It was not obvious whether Bristol had been included in Somerset or Gloucester for the purposes of computing the county average grain prices in the period 1770 to 1820. However, a town-level breakdown of the Corn Returns for 1819 was given in response to a Parliamentary enquiry.40 We were then able to calculate the county averages under the assumptions that: a) Bristol was in Somerset; and b) Bristol was in Gloucester. By comparing these calculations with the printed county returns we were able to establish that Bristol had been categorised as part of Somerset.

39 Ordnance Survey, Ordnance Survey Gazetteer of the British Isles (4th Edition, 1999). 40 British Government, Appendix to Report from Select Committee on Petitions Relating to Agricultural Distress, BPP 1820, vol. 2, 164.

28 29 30 31 32 33 34 References.

Barnes, Donald Grove, A History of the English Corn Laws, 1660-1846 (Routledge, London, 1930).

British Government, ‘Select Committee on the Sale of Corn,’ BPP (1833), vol. 7, 54- 6.

Evans, Eric J., The Contentious Tithe (Routledge and Kegan Paul; London, 1976).

Fay, C. R., ‘Price Control and the Corn Averages under the Corn Laws,’ Economic Journal (1926), vol. 36, 151-2.

Giffen, Robert, ‘The Gazette Average Prices of Corn,’ Journal of the Statistical Society (1879), vol. 42, 709-723.

Jacob, William, Considerations on the Protection Required by British Agriculture and on the Influence on the Price of Corn on Exportable Productions (8 vols., London, 1814).

Jacob, William, Report on the Trade in Foreign Corn, and on the Agriculture of the North of Europe (2nd ed., 8 vols., London, 1826).

Jacob, William, Tracts Relating to the Corn Trade and Corn Laws (8 vols., London, 1828).

Jeandupeax, Ludivine, ‘OCR Packages,’ mimeo., University of Bristol.

Lawes, J. Bennet, and J. H. Gilbert, ‘Report of Experiments on the Growth of Wheat for 20 Years in Succession on the same Land,’ Journal of the Royal Agricultural Society of England (1864), vol. 25, 93-185.

Ordnance Survey, Ordnance Survey Gazetteer of the British Isles (4th Edition, 1999).

Percival, John, Wheat in Great Britain (Duckworth, London, 1934).

Vamplew, Wray, ‘A Grain of Truth: the Nineteenth Century Corn Averages,’ Agricultural History Review (1980), vol. 28, pt. 1, 1-17.

Young, Arthur, The Farmer’s Tour through the Eastern Counties of England (London, 1770).

35